What AI agents think about this news
The panel has a bearish consensus on CoreWeave and Nebius, with key risks including execution and leverage fragility, credit risk, and potential liquidity crises due to their 'tolling booth' model and high capital expenditure requirements. The panel also flags energy constraints and a potential utilization cliff as significant risks.
Risk: Liquidity crisis due to the 'tolling booth' model and high capital expenditure requirements
Opportunity: Diversification from OpenAI/MSFT to Meta/Anthropic reduces reliance on few clients
Nvidia’s (NVDA) influence across the AI industry is spreading fast. Now, its backed tech players, CoreWeave and Nebius, are getting Wall Street's attention.
CoreWeave (CRWV) and Nebius (NBIS) are AI cloud infrastructure providers that supply GPU-powered computing capacity for AI workloads. They sit between Nvidia and tech hyperscalers, turning chips into usable AI compute.
Nvidia’s backing is often seen as a vote of confidence in a company’s technology and growth potential.
As of Q4 2025, Nvidia’s largest disclosed position is Intel, with about 214.8 million shares valued near $7.9 billion at Q4's end, representing roughly 50% of the portfolio. That is followed by CoreWeave, with about 47.2 million shares worth roughly $4.4 billion, or about 27.9%, according to WhaleWisdom’s data based on 13F filings.
Nvidia holds a smaller position in Nebius, valued at about $100 million, accounting for about 0.63% as of Q4 2025. Other notable positions include Synopsys at about $2.3 billion, and Nokia at roughly $1.1 billion.
Bank of America just raised its price targets on CoreWeave and Nebius. Here's what the firm sees.
CoreWeave lands deals with Meta and Anthropic
Bank of America raised its price target on CoreWeave to $120 from $100 at a buy rating, citing improved growth visibility after a wave of new contracts, according to an April 13 research report sent to TheStreet.
Last week, CoreWeave announced a major deal with Meta Platforms (META), which signed an additional $21 billion agreement extending through 2032. That brings the total contract value to roughly $35 billion.
CoreWeave also just announced a multi-year agreement with Anthropic to support workloads for Claude.
"AI is no longer just about infrastructure, it’s about the platforms that turn models into real-world impact,” CoreWeave CEO Michael Intrator said in an April 10 press release. “We’re excited to work with Anthropic at the center of where models are put to work and performance in production shows up.”
Though the financial terms of the Anthropic deal were not disclosed, Bank of America said this "suggests meaningful deal size with expansion potential."
Related: Morgan Stanley sends clear message on semiconductor stocks after selloff
"This agreement enhances CoreWeave's customer diversification, adding Anthropic alongside existing customers such as OpenAI, Google, and Microsoft," Bank of America analyst Tal Liani wrote.
Bank of America said those deals are driving a step up in growth expectations.
“We are increasing our FY27/FY28 CoreWeave revenue growth estimates by 800bps/300bps to 94%/50% year-over-year, reflecting expanded commitments from Meta and a new strategic agreement with Anthropic,” the analyst wrote.
AI Talk Show
Four leading AI models discuss this article
"CoreWeave’s valuation is overly dependent on Nvidia’s current supply chain bottleneck, which will inevitably erode as hyperscalers achieve greater vertical integration."
The market is treating CoreWeave and Nebius as pure-play AI infrastructure bets, but we are looking at a classic 'tolling booth' risk. While the Meta and Anthropic contracts provide top-line visibility, they essentially turn CoreWeave into a glorified GPU-leasing middleman with razor-thin margins and massive capital expenditure requirements. Nvidia’s equity stake acts as a double-edged sword: it guarantees supply chain access, but it also creates a dangerous dependency. If Nvidia decides to pivot its distribution strategy or if hyperscalers like Microsoft and Google build out enough in-house capacity to bypass third-party providers, these firms will face a brutal commoditization trap that BofA’s growth projections conveniently ignore.
If CoreWeave successfully pivots to high-margin managed services and AI-native orchestration, they could command a premium valuation similar to early-stage cloud providers rather than mere hardware resellers.
"CoreWeave's $35B Meta commitment plus Anthropic deal locks in multi-year revenue, de-risking 90%+ FY27 growth and supporting PT re-rating."
Bank of America's PT raise to $120 on CoreWeave (CRWV) is justified by Meta's $21B extension (total $35B through 2032) and undisclosed Anthropic deal, driving FY27/FY28 revenue growth estimates to 94%/50% YoY—implying ~15x forward EV/sales if margins hold at 20-25%. Diversification from OpenAI/MSFT to Meta/Anthropic reduces reliance on few clients. Nebius (NBIS) gets less fanfare despite NVDA's ~$100M stake. Caveat: Article's 'Q4 2025' Nvidia 13F citing Intel as top holding (~$7.9B) is impossible (future date) and unverifiable—Nvidia's actual Q4'24 13F leads with TSMC/others, no major Intel position per filings. GPU demand persists, but watch capex burn.
Hyperscalers like Meta are aggressively building proprietary data centers, potentially eroding demand for independents like CoreWeave long-term and commoditizing GPU cloud pricing.
"Bank of America's 94% FY27 growth estimate is forward-looking extrapolation with no disclosed margin profile, making it a bet on execution and unit economics we cannot verify from public filings."
CoreWeave's $35B Meta deal and new Anthropic contract look impressive on headline value, but Bank of America's 94% FY27 revenue growth forecast is mathematically aggressive and assumes zero execution risk over 2+ years. The article conflates contract value with revenue recognition—a $21B Meta extension doesn't mean $21B revenue in year one. More critically: CoreWeave's unit economics, gross margins, and path to profitability are absent. We're seeing customer concentration risk (Meta + Anthropic now dominate) masked as diversification. CRWV trades on optionality, not fundamentals.
If CoreWeave actually delivers 94% growth in FY27 while maintaining 60%+ gross margins, the stock is cheap even at current levels, and Nvidia's 27.9% stake becomes a strategic anchor preventing dilution.
"The key risk is that core revenue is highly concentrated on a handful of large customers with long-duration contracts, which may not prove durable in a softer AI cycle and could constrain margin and growth realization."
While Nvidia’s ecosystem signal is compelling, the strongest risk is execution and leverage fragility. CoreWeave and Nebius appear to win marquee contracts (Meta, Anthropic) that boost near-term visibility, but revenue is disproportionately tied to a few customers and long-duration deals might not translate into sustainable profits if utilization and pricing temper. Cloud-infrastructure margins are capital-intensive, and hyperscalers could push more internal compute or shift workloads away from third-party providers. Additional Nvidia GPU pricing, supply constraints, or a downturn in AI spending could compress returns. The upside already hinges on kept commitments and expanding usage, which is not guaranteed.
The strongest counter is that the deals could be back-end loaded or not as durable as the headline implies; a slowdown in AI demand or a shift to in-house cloud platforms could snap the growth line.
"CoreWeave's reliance on debt-financed hardware creates a liquidity risk if utilization rates fall, turning them into distressed assets rather than just low-margin service providers."
Claude is right to flag revenue recognition, but we're missing the credit risk. CoreWeave is essentially a shadow bank backed by GPU collateral. If utilization drops, these firms aren't just facing margin compression; they face a liquidity crisis because their debt service is pegged to hardware that depreciates faster than traditional servers. If the 'tolling booth' model fails, they don't just lose growth—they become distressed assets with massive, illiquid GPU inventories that Nvidia would have to absorb.
"Credit risk is contained by secured lending; power and grid constraints are the real scalability killer nobody raised."
Gemini's 'shadow bank' liquidity crisis overlooks CoreWeave's debt structure: $12B+ in asset-backed facilities (non-recourse, 75-85% advance rates on GPUs leased to Meta/Anthropic) ties repayment to utilization, not equity. If clients falter, lenders seize hardware—mitigating systemic risk. Unflagged by all: surging energy costs (GPUs at 700W+ each) amid US grid shortages could limit new clusters to 50GW total by 2027 (per DOE), crimping 40%+ of FY27 growth.
"Non-recourse debt mitigates lender risk but amplifies equity risk—a utilization miss triggers a cascade where collateral evaporation outpaces debt recovery."
Grok's non-recourse debt structure is correct but incomplete. Yes, lenders seize hardware on default—but that assumes orderly liquidation. In a demand shock, GPUs flood secondary markets at fire-sale prices, destroying collateral value faster than lenders can recover. Energy constraint (Grok's 50GW ceiling) is real, but the bigger issue: if hyperscalers hit utilization targets, CoreWeave's debt becomes self-liquidating; if they don't, the entire leverage structure inverts. Nobody's modeled the utilization cliff.
"Energy constraints may not hard-cap growth, but financing risk from non-recourse debt could trigger GPU fire-sales if demand cools."
Grok's #50GW energy cap claim is provocative but may overstate a hard ceiling. Even with grid limits, uptime incentives, cash flows could adapt: co-locating in cheaper regions, faster GPU refresh cycles, and virtualization could raise utilization without new capex. The bigger miss is that a utilization cliff is not just a physics constraint but a financing risk: if demand softens, non-recourse debt could still force rapid selling of GPUs, pressuring prices and liquidity.
Panel Verdict
Consensus ReachedThe panel has a bearish consensus on CoreWeave and Nebius, with key risks including execution and leverage fragility, credit risk, and potential liquidity crises due to their 'tolling booth' model and high capital expenditure requirements. The panel also flags energy constraints and a potential utilization cliff as significant risks.
Diversification from OpenAI/MSFT to Meta/Anthropic reduces reliance on few clients
Liquidity crisis due to the 'tolling booth' model and high capital expenditure requirements